HDMI Color Processing Board Used As An FPGA Dev Board To Mine Bitcoins


The blue board seen above is the guts of a product called the eeColor Color3. It was designed to act as a pass-through between your television and HDMI source device. It boasts the ability to adjust the color saturation to suit any viewing conditions. But [Taylor Killian] could care less about what the thing was made for, he tore it open and used the FPGA inside for his own purposes.

The obvious problem with this compared to a proper dev board is that the pins are not all broken out in a user-friendly way. But he got his hands on it for free after a mail-in-rebate (you might find one online for less than $10 if you’re lucky) and it’s got an Altera Cyclone IV chip with 30k (EP4CE30F23C6N) gates in it so he’s not complaining. The first project he took on with his new toy was to load up an open source Bitcoin mining program. The image above shows it grinding away at 15 megahashes per second while consuming only 2.5 watts. Not bad. Now he just needs to make a modular rack to hold a mining farm.

73 thoughts on “HDMI Color Processing Board Used As An FPGA Dev Board To Mine Bitcoins

    1. So repurposing a pass-through graphics adjuster to run your own software isn’t a hack? No one’s forcing you to mine or accept bitcoins as a currency. I’m sure they could put up something with an Arduino if you’d like.

    2. The whole mining thing is a good argument against the currency to me*, so its hardly pushing bitcoins to mention it.
      This is still a nice hack, mind.

      *I mean, honestly, wasting cpu power (and thus real power) to make something which amazon or google could devalue hugely overnight if they diverted there own servers to it.

        1. This little guy is actually rather bad at mining also.
          2.5 Mhash is something that an average dual core can push out.
          I have seen custom asic chips that can push out 2.5 Ghash at the same wattage.
          Ofc, its not as cool.

          1. (cant reply to a comment that deep)
            @ehrichweiss I dont think you are running a very efficient miner then.
            Also 2.5 Mhash for a gpu is rather terrible. modern day graphics cards can push out over 500 Mhash, but ofcourse, for 200 watts.

      1. Bitcoin doesn’t work that way.

        The block creation (i.e. “mining”) rate of the entire Bitcoin network is constant, no matter how much computing power is present – the problem gets more difficult with more computing power. So unless they’re planning a 51% attack (which is a whole ‘nother thing and rather unlikely), all they can do is make mining more difficult for individual miners.

        1. Which is the exact same as making your own worth more isnt it?
          If they have a large share of the server power, they have a larger share of whats produced…..even if, in absolute terms, the total mined is the same.

          1. not the same. the quantity of bitcoins grows at a constant rate. the value is determined based on supply and demand. It does not depend on who holds the bitcoins. so, Amazon could dedicate enough resources so that they are the likely party to solve the hash and be awarded the new bitcoins instead of someone else, but they cannot cause the supply to increase at a faster rate than it would have increased in their absence.

          2. dave: no. The quantity of new mined bitcoins decreases with time.

            The TOTAL amount of bitcoins is known and fixed. Mining is more and difficult.

            Think of a gold mine.

      2. Amazon and Google are not set up to do this, as mining specific ASIC hardware is now available, and Amazon/Google’s servers can’t possibly compete with that.

        Even if they could, directing a ton of computation power at the Bitcoin network does not increase the long term rate of supply. The mining difficulty self-adjusts every 2016 blocks mined, to maintain a rate of 25 bitcoins per 10 minutes. Besides, ultimately only 21 million coins can be mined, no matter how much compute power you spend on mining, and 11 million are already mined.

        1. if HDCP is supported by this thing, then the transcoder chips sits between the hdmi plug and the fpga. The secret keys are not managed by the fpga itself.

  1. Has anyone mined Bitcoin on the “Digilent Spartan-3 Starter Board”? I’m lazy and I don’t know VHDL. (I should learn, I suppose, since I own an FPGA board.)

    1. Yes, I like it too, it fits well with, refilling inkjet cartridges, expanding the features on a Fluke DVM or whats-its-name oscilloscope, turning a child’s electronic toy into a Linux machine, its a nice David v. Goliath scenario, turning tables on companies that follow the Gillette paradigm of “giving away the razor and making a huge profits on the blades”.
      Or anything related to $ony. B^)

        1. I’ll give you that the refill heads are finely engineered tools, a pack of double edged safety blades is a couple bucks at CVS. The blades aren’t rocket science.

          1. The Tool & Die I used to work at made a die to stamp out scaple blades once. It’s trickier than it sounds. The idea of how it works is simple enough, you coin the cutting edge inbetween two anvils ground to the appropriate angle. Unfortunetely, you need near zero clearance between them in a multi-ton high speed press. Did I mention how brittle Tungeston Carbide is…

            We also made stuff for the US military that went in warheads of missles. Real rocket science

          2. To elaborate on this a little more, I’ll take the example of the old double edged razor blade for a butterfly razor. The material is around 0.004″ thick (0.102mm) Without knowing the shear strength of the steel used I’ll go with a standard 5% clearance between punch and die, that equates to 0.0002″(0.005 mm). What happens is the punch enters the material being stamped, plastic deformation untill the shear strength is exceeded, then it fractures. Ideally the die size should be the same size as the fractured material being forced through, You end up wth a straight section and the break tapered larger looking at the cut. Too much clearance and it will drag the stock down with it and tear, you end up with a burr on the downward side, too little clearance and the force needed from the press goes up and could possibly crack the die block and punch.
            The pilots and leader pins for the die should have a clearance less than half that, If the press has slop, or the die and punchplate get out of alignment any more than that you have shattered a lot of expensive tool-steel and carbide and production will have to stop untill a replacement tool can be built. Shouldn’t ever really happen, that’s what the parabolic points on the pilots and leader pins/ TR Jones are for.

            As far as the coining of the cutting edge, I think the idea was to increase the life of the grinding wheels, diamond grinding wheels are expensive and they do wear out.

  2. Bravo!! Just snagged 2 for $15 apiece on eBay from diff sellers, still a few left last I checked. Have a “prototype” board from Altera with only DVI in but have wanted to dabble with manipulating/compositing HDMI signals on an FPGA and this looks like it *may* work for that.

      1. Correct on the cheaper. However, for the pure and simple fact that it has a /shedload/ of I/O, and a suite of on-board goodies (Accelerometer and DAC in addition to what’s on the Color3), it might be somewhat more appealing if you want to do anything hardware-based.

        I’ve got a DE0 somewhere at the office. Something of an impulse buy for an upcoming Uni module next year. Does anyone know of any decent “FPGAs for electrically-proficient Dummies” sort of guides? Am itching to jump into this area!

    1. It is all a waste of money ,because 15 MH/s is not going to pull you out of bed.

      ROI is going to be about 382 days at current difficulty& price and that is not accounting for the time he spent reverse engineering an under-powerd PCB.

        1. My wife ‘corrected’ me the other day when I said it this way. I said “Yes, I COULD, indeed, care less. But I don’t because it would require more energy than caring as little as I already do.” :D

          1. “Couldn’t care less” might be grammatical, but it’s still illogical.

            If the speaker’s level of caring is exactly zero, then the statement is a contradiction; if you cared enough to comment on it then obviously your level of caring is non-zero.

            If the speaker’s level of caring is non-zero and positive, then by definition they could care less, because the well-ordering theorem guarantees the existence of an even smaller level of caring. Therefore the statement is still contradictory.

          2. ““ Couldn’t care less”” is a exaggeration – but its what people mean. It at least infers a very very low level. A level so low they cant conceive of lower.

            “Could care less” is utterly meaningless, as the level of carrying is anywhere between infinitesimal, and infinite.

          3. @Joe

            This was a new one! Didn’t think people could come up with more bad examples why the “couldn’t”-version is wrong.

            The thing is, we live in world where just standing silent in a conversation is rude, and also confusing. Therefore, the sentence is logical to use. Sorry to break your bubble.

      1. If you remove the meaning of a sentence, then you didn’t shorten it correctly. It’s still pretty short if you add “n’t”, and then it would still be logical.

    1. are those little DC-DC modules really rated at 50A continuous?
      seeing how $10-20 12V 45-65A IBM/DELL/HP server power supplies RC people use to charge lipo packs are the size of and arm and still get quite hot I have a hard time believing that.

    1. There’s a few more factors that go into this. These look to be about $20 everywhere now since most of the cheap/free ones have been snatched up. Also there is limited stock as it’s a discontinued/failed product. So good luck reaching those figures with even a 20K investment. Still a nice board for the price though, perfect if you need FPGA/HDMI.

    1. Personally I Have a box full of brand new Kodak PalmPix cameras for Palmpilots ($1 each)
      They got 512KB dram, AT90S8615 and altera flex epf8282atc100-4
      Sadly I wasnt able to find Quartus version supporting this chip (free/education versions lack license)

      I also have Pinnacle Studio MP10 with EPF6016QC208 (same story about Quartus).

      Those are from the times when FPGA was magic and you had to pay out of your ass for dev tools :(

  3. I looked these up on amazon, and they were about $25. Then I took a nap, and when I woke up they were at $45. I hate amazon’s dynamic pricing sometimes…

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.